Missing value analysis revealed that there were 130 complete cases and 23 incomplete cases. Of these, one case had more than 20% of items missing and was removed from the data set (Peng, Harwell, Liou, & Ehman, 2006). Data were missing completely at random for the re- maining incomplete cases (Little’s 2 662.51, df 675, p .63). As such, missing values were imputed using the expectation maximiza- tion algorithm at the variable level (Cole, 2008). Following imputation, in accordance with the recommendations of Osborne (2013), univariate and multivariate outliers (p .001) were re- moved from the data set (N 4). Although this process resulted in data that were approximately univariate normal, estimates of multivariate kurtosis (Mardia’ s normalized coefficient 90.98) indicated that the data remained multi- variate asymmetrical. Therefore, a bootstrap- ping procedure that drew 5,000 replication sam- ples with replacement was used. Confidence intervals (CIs) and p values associated with the correlation and regression coefficients are those derived from the standard errors from this boot- strapping procedure. The data screening and cleaning procedure yielded a final sample of 148 (89 men, 59 women; Mage 15.12 years, SD 1.64).